An Interaction Coefficient Control Method for Setting Initial Solutions to Ising Machines
Soma Kawakami, Kentaro Ohno, Dema Ba, Satoshi Yagi, Junji Teramoto, Nozomu Togawa · 2024
Ising machines, such as quantum annealers and coherent Ising machines, can efficiently search for optimal or sub-optimal solutions of difficult combinatorial optimization problems. At that time, if a tentative solution initially generated is input to an Ising machine, a better solution can be obtained. However, it cannot directly be input to some types of Ising machines, due to their computational principle. This paper proposes an interaction coefficient control method for setting initial solutions to a wider type of Ising machines. When an initial solution to an original combinatorial optimization problem is given, the proposed method efficiently transforms it to interaction coefficients and combines it into an original Ising model. The combined Ising model can be efficiently solved by any type of Ising machines. We have demonstrated the efficiency of the proposed method through the experimental evaluations.